Canada logs 74 AI hallucination cases in 6 months, 10x rise since 2024
Canadian courts have documented 74 decisions involving AI-fabricated legal citations in the first six months of 2026, according to data from Courtready, a legal-technology platform and access-to-justice watchdog. The figure represents a sharp acceleration: 87 total cases in 2025 and just 7 in 2024, prompting legal experts to characterize the documented incidents as the "tip of the iceberg" [CTV News].
Scope and Scale of the Problem
The 74 H1 2026 cases span 57 distinct courts and tribunals across Canada, with self-represented litigants accounting for 81% of incidents, according to Courtready. However, approximately 19% of cases—roughly 30 legal professionals—involved licensed lawyers who have faced sanctions or regulatory criticism from provincial law societies and bench officers. The data snapshot captures decisions made as of July 13, 2026, but Courtready's full database now records 185 total instances of AI-related judicial misconduct, including cases beyond fabricated citations.
Largest Penalty: Ontario Lawyer Sanctioned $31,150
The highest cost award to date came in *Mazaheri v Law Society of Ontario*, a 2026 Ontario Law Society Tribunal decision imposing $31,150 CAD in costs against suspended lawyer Shahryar Mazaheri. The tribunal found that Mazaheri had filed legal submissions containing fabricated, non-existent case citations generated using the AI tool Grok without personal verification. In its ruling, the Law Society Tribunal of Ontario stated: "The verification duty attaches when a lawyer signs and files. Using generative AI without personally checking every citation is sanctionable, can trigger a personal cost award, and can lead to a law society or regulatory referral." Similar findings were reinforced in an Alberta appeal ruling emphasizing that the duty to verify citations cannot be delegated to machines.
Regulatory Response and New Requirements
The Law Society of Ontario has imposed adverse cost awards, formal findings, and regulatory proceedings against sanctioned practitioners. The Federal Court now requires mandatory disclosure of AI use in all filed documents, and multiple provincial jurisdictions have issued guidance requiring lawyers to personally verify every citation before submission. Courts have also begun tracking which generative AI tools (Grok, ChatGPT, Claude) generated problematic material, establishing an audit trail for disciplinary proceedings.
The cascade of cases underscores a structural vulnerability: agents and AI systems generating legal filings at scale without human-in-the-loop verification. As the legal system continues documenting these incidents, experts warn that visible sanctions represent only flagged cases, while many hallucinated citations may have influenced litigation outcomes before discovery.
